American high school seniors posted historically low scores on the nation’s most authoritative academic exam in 2024, extending a decline that began well before the pandemic. The results, drawn from the National Assessment of Educational Progress conducted between January and March 2024, show fewer 12th graders reaching basic proficiency in both reading and math. The slide tracks closely with the era of mass device adoption in classrooms, and the rapid spread of AI-powered tools now threatens to widen the gap between students who build real skills and those who learn to outsource thinking to a chatbot.
A Decade of Decline, Then a Cliff
The trouble did not start with COVID-19 school closures. Federal data from the NAEP long-term trend assessment shows U.S. student achievement was already sliding after 2012, years before remote learning entered the picture. The pandemic then accelerated the drop. Reading and math scores for 13-year-olds fell sharply between 2020 and 2023, according to the same series, which tracks performance at ages 9, 13, and 17.
For 12th graders specifically, the 2024 results are the worst on record. The NCES reading assessment documents a post-2019 decline that pushed scores to their lowest point since the test began in 1992. Math tells a similar story: the grade 12 mathematics results show continued erosion compared to both 2019 and earlier trend points. “These students are taking their next steps in life with fewer skills,” NCES leadership said of the findings, as reported by the Associated Press.
That quote captures the real cost. These are not abstract scale-score movements. The NAEP’s item-level analysis connects score ranges to specific abilities: at lower percentiles, students struggle to summarize a passage or solve problems that require more than one step. A generation of seniors is entering college and the workforce measurably less prepared than their predecessors were a decade ago.
What Screens Did Before AI Arrived
The timeline of score declines overlaps almost exactly with the period in which U.S. schools embraced one-to-one device programs and digital curricula. Between roughly 2012 and 2019, districts spent billions equipping students with Chromebooks, iPads, and online learning platforms. Achievement did not improve. It fell.
International data reinforces the pattern. The OECD’s PISA 2022 assessment found that performance among 15-year-olds declined sharply between 2018 and 2022 across dozens of education systems. The same study includes survey-linked findings on self-reported digital-device distraction in class, drawing a direct line between screen time during instruction and weaker outcomes. Students who reported frequent distraction by devices scored lower, and the effect held across countries.
The OECD’s U.S. country note shows American students’ reading, math, and science performance declining relative to both 2018 and 2012 benchmarks. The drop was not unique to the United States, but the country’s slide was notable given its level of per-pupil spending and technology investment.
The dominant assumption in education policy for the past decade has been that more technology equals better learning. The data says otherwise. Devices were treated as a delivery mechanism without enough attention to whether the content and pedagogy running on them actually built durable skills. Schools digitized instruction but did not always redesign it, and the result was a generation of students toggling between a math lesson and a group chat on the same screen.
Skill Erosion at Every Level
The damage is not confined to one age group or one subject. The NAEP long-term trend series, which uses a distinct methodology from the main NAEP to allow comparisons stretching back decades, shows declines at ages 9, 13, and 17. The 2024 grade 12 results add the most recent data point to that arc.
Distributional breakdowns tell an even sharper story. The national trends report for grade 12 reading displays percentile and achievement-level data from 1992 through 2024. The losses are steepest at the bottom of the distribution, meaning the students who were already struggling fell furthest behind. That pattern matters because it signals widening inequality: higher-performing students held relatively steady while their lower-scoring peers lost ground.
When fewer students can meet even the “basic” threshold on a federal exam, the consequences ripple outward. Employers report difficulty finding entry-level workers who can write clearly or handle quantitative tasks. Community colleges absorb the cost through remedial coursework that delays degree completion. The NAEP numbers are not just an education story. They are a workforce readiness problem with direct economic consequences.
AI Tools Risk Compounding the Problem
Into this weakened foundation, generative AI has arrived at speed. ChatGPT launched in late 2022, and by the 2023-2024 school year, students across the country were using large language models to draft essays, solve homework problems, and even complete take-home exams. For teenagers who already struggle with reading comprehension or multi-step math, the temptation to let a chatbot do the heavy lifting is obvious.
Educators now face a double bind. On one hand, AI can act as a powerful scaffold: a student stuck on a calculus problem can ask for a hint, or a reluctant writer can get help outlining a paragraph. On the other hand, the same tools make it easy to bypass the struggle that builds fluency. A senior who has never had to wrestle with revising a rough draft may arrive in college without the stamina or skill to write independently under time pressure.
The risk is greatest for students already on the margins. Well-supported teenagers in high-resource schools are more likely to be taught how to use AI as a supplement (checking work, generating practice questions, or exploring alternative explanations). Their peers in under-resourced settings may encounter AI primarily as a shortcut, with little guidance on when to turn it off and think for themselves. That dynamic threatens to mirror the earlier device wave, in which technology widened gaps instead of closing them.
There is also a measurement problem. The NAEP, SAT, and state exams still rely heavily on proctored, closed-book testing. If day-to-day coursework becomes saturated with AI assistance, the gap between what grades suggest and what students can do unaided will grow. The 2024 NAEP results, already the lowest in decades, could be a floor rather than a ceiling if classroom practice drifts further from the independent performance those exams are designed to capture.
Redesigning Learning for a Hybrid World
The response cannot be to ban AI or rip devices out of classrooms. Digital tools are now woven into higher education and most workplaces; students must learn to navigate them. The question is how to design learning so that technology amplifies, rather than replaces, core skills.
One starting point is to separate practice from production. During key parts of a lesson (reading a complex text, working through problem sets, or drafting in-class essays), students can be asked to work without AI or internet access, preserving a space where teachers can see what they truly know. Later, they can bring in AI to revise, extend, or critique their own work, turning the tool into a mirror rather than a crutch.
Assessment also needs to evolve. More frequent low-stakes quizzes, oral examinations, and in-class writing can help teachers detect when students are leaning too heavily on outside help. At the system level, policymakers can use trend data from instruments like the national math assessment and the grade 12 reading exam to monitor whether reforms are actually improving independent performance.
Teacher training is another pressure point. Many educators were handed devices and software with minimal preparation; the same mistake cannot be repeated with AI. Professional development that models how to design AI-resistant tasks, how to teach students about academic integrity in an AI era, and how to use these tools for their own planning and feedback could turn a perceived threat into a practical ally.
A Narrowing Window
The long-term trend charts are a warning that extends beyond any single technology cycle. The decade-plus slide in achievement, now capped by record-low 12th-grade scores, suggests that the basic contract of American schooling (send your child to class and they will emerge ready for adult life) is under strain. Generative AI did not cause that erosion, but it could lock it in if schools allow tools to substitute for effort.
There is still time to choose a different path. The same systems that delivered widespread device access can be used to push out high-quality curricula, tutor students who fell behind during the pandemic, and give teachers better data on who is struggling. AI, deployed thoughtfully, can help personalize practice and free up human educators to focus on the complex, relational work that no chatbot can do.
The 2024 NAEP results are a snapshot of where American students stand at the dawn of the AI age: more connected than ever, but less prepared on the fundamentals. Whether the next decade of data looks better will depend on whether schools treat technology as a shortcut around learning or as a demanding partner in the hard work of mastering it.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
